50 Volt 3 A power supply

[Olegkuskhov]

Hi. I plan to build a linear power supply that capable of adjusting the voltage from let say 2 to 50 volt and maximum current of 3 amp at any output value. I dont have any strict requirement like fast response or current limiting feature. I just want that if the current goes above 3 amp, the psu will be protected.

I Search online and find that people mostly do it with lm 723 or go with more tricky approach using op amp to control the regulation loop. Can someone reccomend for me any working design, and i prefer using tip 3055 as i have like 16 of them lying around. The size and weight is not really a problem for me (big heatsinks, etc). Thanks in advance

[Olegkuskhov]

Hi thanks a lot for the circuit. Appreciate your work. Definetly will try to undesrtand it first and try it.

[coromonadalix]

lm723  can do higher   

but you need to make it work in a "floating voltage" configuration, see its datasheet

meaning the lm723 supply line is separated from the main who's higher,

723 can not go higher the 40v on its vcc input pin

you have many lm723 designs "floating" on google ....

[dobsonr741]

A 50V 3A supply, without a multi tap transformer or other ways of reducing the input voltage to the pass transistor is not practical.

Think about the power dissipation at 2V, 3A setting: 150W on the pass transistor(s). That is a huge heat sink with forced air cooling.

If you attempt it, add thermal fuse at least.

[naiclub]

lm723  can do higher   

but you need to make it work in a "floating voltage" configuration, see its datasheet

meaning the lm723 supply line is separated from the main who's higher,

723 can not go higher the 40v on its vcc input pin

you have many lm723 designs "floating" on google ....

If you can do that, it's considered very good. I think I did it once, many years ago. But it doesn't answer the question. It doesn't have negative light, only + and G. If you answer that you can't do it, add 2 pieces to the rank. That way it is considered that G is not connected together.

[naiclub]

A 50V 3A supply, without a multi tap transformer or other ways of reducing the input voltage to the pass transistor is not practical.

Think about the power dissipation at 2V, 3A setting: 150W on the pass transistor(s). That is a huge heat sink with forced air cooling.

If you attempt it, add thermal fuse at least.

That said, it is true. Which is very hot, the temperature may touch 100 degrees. So I solved the problem with 2 large fans. Which helped a lot.

[Kleinstein]

Simple cooling is not enough. Especially with a higher voltage as 50 V, there is also a SOA limit due to 2nd breakdown.
So at 60 V CE voltage the 2N3055 can no longer allow some 115 W of thermal loss as shown as ptot, but less.  Chances are the raw voltage before the regulator can even exceed 60 V at low load.
The ON data-sheets give an safe current of some 0.7 A with a 60 V drop. With a Tcase of more than 25 C this would be even less.
Chances are the TIP3055 is not better than the TO3 version.

For the high power one would need multiple transistors in parallel and maybe more power full ones than the TIP3055. For the TIP3055 it would be at least some 5 transistors in parallel, maybe better 10, as current sharing may not be perfect and the heatsink likely also not.

A high power linear supply is not an easy or beginners project. For the start one should first start smaller (e.g. 30 V and 1A) - already this has it's challanges.

[reboots]

The TL783 high-voltage linear regulator has a maximum input-output differential of 125V. Unfortunately this part is being discontinued, and is in last time buy status. You would still need external pass transistors for your desired current range, and the caveats mentioned about power dissipation would still apply. The datasheet has a few current boost application examples, using obsolete transistors.

https://www.ti.com/product/TL783

[naiclub]

Or you will do this Which is quite difficult because
2 sided pcb but has complete POTECH system

[naiclub]

The TL783 high-voltage linear regulator has a maximum input-output differential of 125V. Unfortunately this part is being discontinued, and is in last time buy status. You would still need external pass transistors for your desired current range, and the caveats mentioned about power dissipation would still apply. The datasheet has a few current boost application examples, using obsolete transistors.

https://www.ti.com/product/TL783

I've made this before and it got really hot.

[Jwillis]

I found it to be much easier to make high voltage high current stable variable power supplies with high voltage op amps and completely separating the bulk of the current from the control section. Essentially the control board is powered separately from the high current supply. This virtually eliminates the effects of ripple on the voltage and current control. I currently working on a 0 to 40V 0 to 20A variable supply that is stable and ripple free that is capable of up to 70V simply by changing the transformer. By changing up the op amps and transformer it could be capable 80 to 85V.
3 amps would be very easy at 60V. The main thing is selecting the transformer. If you want 60V, the transformer will need to be at the very least 65 VAC RMS but 70VAC RMS would be better. This depends a lot on the transformers own regulation. Increasing the VA won't help. Don't rely on the rectified voltage as a basis to your final output voltage either. Always base your output power on the RMS of the transformer. Use the rectified filtered voltage to determine the working voltage of the components used.
The transformer will be the most expensive part of your power supply, so it's wise to get it right the first time. This includes the VA , the AC RMS voltage and if you intend on having multiple taps for your power transistors and other secondary taps for peripherals. 

[alligatorblues]

I am against building rectifiers as a DIY project. WHY? SMPS have too much noise, although their pretty easy to build. Linear rectifiers are a combination of art and Zen. The idea of analog electronics is quiet. Once the noise is there, it's tough to get it out. They are not at all easy to build. The BOM has about 300 items, not including chassis and PCBs. There are infinitely more ways to fail than to succeed. But it's like anything else, the more time, effort and money you invest, the better the result.

60V is a bit much for a typical hobbyist. And 3A is a bit low. A 0-40V 0-5A would cover 95% of all projects. One of the difficult tasks is the readout. Definitely LED 7-segment down to mV and mA resolution. Tolerance is good at < 1 mV. And, Insurance is important, because home brew rectifiers ruin more expensive electronics than any other single thing in the history of creation!

And, for the time you put in, the parts you purchase, the boards you design, the endless hours of drudgery while friends are out having a blast, and not the remotest hope of any gain of any kind, it's always better to purchase a DC rectifier, or power supply (PSU). EBay has 200,000+ DC power supplies for sale presently. Lambda and Sorensen are the best major suppliers.

You can probably pick up a 0-40V 0-5A linear unit for $150 + $40 shipping. Have a look. And for a DIY project, choose something that has some potential for gain. One thing that's in big demand is highly accurate digital humidity gauges with temperature sensors, and basic functions like calculating dew point, of course an A/D converter to drive a digital readout.

There are humidity sensors with RTD temp sensors built in. That would be worth building, because such devices are in high demand, and sensors to make it feasible are a fairly recent development. The applications are endless. Industrial and scientific monitoring. Portable access to that information. You can run the entire device on 5VDC, so all it needs is a USB connector and smart phone reserve power brick.

So why don't I just do it if it's such a good idea. I have several inventions I hold the patent on, and I license use of them, or parts thereof. I have an electrical and chemical analytical lab. I'm designing a deep learning neural network AI rig. So, I don't want to do that too. But if you could design a linear rectifier that's 0-60V and 0-3A, you can definitely do the digital humidity &temp meter.         

[naiclub]

I am against building rectifiers as a DIY project. WHY? SMPS have too much noise, although their pretty easy to build. Linear rectifiers are a combination of art and Zen. The idea of analog electronics is quiet. Once the noise is there, it's tough to get it out. They are not at all easy to build. The BOM has about 300 items, not including chassis and PCBs. There are infinitely more ways to fail than to succeed. But it's like anything else, the more time, effort and money you invest, the better the result.

60V is a bit much for a typical hobbyist. And 3A is a bit low. A 0-40V 0-5A would cover 95% of all projects. One of the difficult tasks is the readout. Definitely LED 7-segment down to mV and mA resolution. Tolerance is good at < 1 mV. And, Insurance is important, because home brew rectifiers ruin more expensive electronics than any other single thing in the history of creation!

And, for the time you put in, the parts you purchase, the boards you design, the endless hours of drudgery while friends are out having a blast, and not the remotest hope of any gain of any kind, it's always better to purchase a DC rectifier, or power supply (PSU). EBay has 200,000+ DC power supplies for sale presently. Lambda and Sorensen are the best major suppliers.

You can probably pick up a 0-40V 0-5A linear unit for $150 + $40 shipping. Have a look. And for a DIY project, choose something that has some potential for gain. One thing that's in big demand is highly accurate digital humidity gauges with temperature sensors, and basic functions like calculating dew point, of course an A/D converter to drive a digital readout.

There are humidity sensors with RTD temp sensors built in. That would be worth building, because such devices are in high demand, and sensors to make it feasible are a fairly recent development. The applications are endless. Industrial and scientific monitoring. Portable access to that information. You can run the entire device on 5VDC, so all it needs is a USB connector and smart phone reserve power brick.

So why don't I just do it if it's such a good idea. I have several inventions I hold the patent on, and I license use of them, or parts thereof. I have an electrical and chemical analytical lab. I'm designing a deep learning neural network AI rig. So, I don't want to do that too. But if you could design a linear rectifier that's 0-60V and 0-3A, you can definitely do the digital humidity &temp meter.       

I've been using it for a long time, probably a year. It works well with no problems.
And I already have some spare parts. So we designed and created it. To meet the need to repair various types of amplifiers, its highlight is a 3-way power supply, such as 50+ 0 50-. For repairs and expansions that are removed from the machine for inspection, we use the power supply that was created. Without having to put it back in the machine before trying to listen to the sound. Most importantly, I can supply high current up to 10A. I use bipolar number 2SC5200 and 2SA1943 in parallel, 5 on each side, making it able to supply high current.

[David Hess]

For the high power one would need multiple transistors in parallel and maybe more power full ones than the TIP3055. For the TIP3055 it would be at least some 5 transistors in parallel, maybe better 10, as current sharing may not be perfect and the heatsink likely also not.

I would use at least 8, so 18.75 watts per TIP3055, but there is another 1 or 2 in there for the driver(s).  Fewer is feasible, but I think 4 is marginal, and layout considerations make 8 as easy as 6.  Foldback current limiting would considerably improve the situation.  Given the already high effort, I would also go to the trouble of adjusting the safe operating area protection for temperature.

I might replace the drivers with a 337 or two and rely on them for the thermal protection, but testing for stability should be done before committing to this, and nobody but me seems to like this sort of design.

60V is a bit much for a typical hobbyist. And 3A is a bit low. A 0-40V 0-5A would cover 95% of all projects.

I usually rely on a separate fixed high current supply for projects that require it instead of a variable output high current supply.

Given my fondness for bipolar tracking supplies, I would also consider this type which would split the power side into two at half of the voltage, which has the advantage of allowing the control circuity to operate without leveling shifting or floating operation, however it also requires doubling the control circuitry and high current bipolar tracking supplies are of questionable value except when designing power amplifiers which have split supplies, and even there a variac is usually a better way; power amplifiers should not require regulated supplies.

[Kleinstein]

With the high power one would like to have some sort of tap switching to reduce the thermal load and required heat sink. The classical way is relay switching at the AC side, like what most cheaper lab supplies do. To keep the current spikes reasonable this likes not too large steps in the voltage and thus more like 2 or 3 taps.
An alternative to relay tap switching is using 2 transistors in series, similar to a class H amplifier. For the floating regulator type circuit the electronic cross over is surprosingly simple. This gives fast cross over between the taps and thus also helps with the SOA limitation and can even include the ripple voltage. With only half the voltage one can get away with less transistors, or at least not needing extra transistors. As an advantage one could get away with a more common split secondary transformer (e.g. 2 x 24 V or so).  The electronic cross over is a bit more effective than relay tap switching that needs some extra reserve for ripple and transients.

Using regulator chips like LM317 / LM337 as drivers is not giving much thermal protection, as the critical heat sink temperature is lower, more like 60 C and not some 120-150 C for the regulator internal protection to kick in. So thermal protection would better be separate.
A foldback current limit can really help with the SOA, but it makes it a bit more tricky to get an accurate current limit.

[David Hess]

Using regulator chips like LM317 / LM337 as drivers is not giving much thermal protection, as the critical heat sink temperature is lower, more like 60 C and not some 120-150 C for the regulator internal protection to kick in. So thermal protection would better be separate.

That is not how thermal protection using an LM317/LM337 is implemented.

The external network determines the current ratio between the output transistors and the LM317/LM337 driver, which may mounted to the same heat sink for thermal tracking.  But the LM317/LM337 mounting is made to deliberately have a controlled but high thermal resistance, so the internal junction temperature of the LM317/LM337 tracks the heat sink, but at a higher level determined by the driver's power dissipation and junction-to-heat-sink thermal resistance.

The original implementation used separate heat sinks as shown below, but either way works.

[naiclub]

lm317 lm337 I've done this before, but it wasn't very good. It can supply not much voltage from 1.25V to 30V by paralleling bipolar transistors, 3 sets on each side, but the current is much higher. Should be more than 10 amps (but IC 317T likes to waste time when I supply too much current), which doesn't answer much of the problem. So I looked for a new circuit that could supply higher voltage. Thus is the origin of my latest cycle.

[naiclub]

lm723, separate power type, floating
https://www.eevblog.com/forum/projects/2x-lm723-max-safe-voltage-for-this-circuit/

[Kleinstein]

Using the LM317/337 as a regulator is tricky. One point is the limited voltage, which makes it hard for a 50 V supply. The other points is that the frequency response is not that well defined and with the extra transistors to take the bulk of the load it is not clear if the circuit is stable.  There are multiple manufacturers for the chip and not all may behave exactly the same in this cicuit.

[naiclub]

1-40 Volt, 50 mA – 5A This circuit should be easy to do.

[David Hess]

Using the LM317/337 as a regulator is tricky. One point is the limited voltage, which makes it hard for a 50 V supply. The other points is that the frequency response is not that well defined and with the extra transistors to take the bulk of the load it is not clear if the circuit is stable.  There are multiple manufacturers for the chip and not all may behave exactly the same in this cicuit.

I have always had good results but my selection of different 317s and 337s has been limited to premium brands and I always used slow output transistors; things might be different with an array of D44VH10s and D45VH10s fast power transistors.

I forgot about the higher voltage requirement.  I wonder what the best way would be to handle it.  My proposal to double up the control circuits to make a dual tracking power supply completely solves it, but would not work at even higher voltages and is complicated unless a dual tracking power supply is an objective.

All of the solutions I found involve a cascode, which makes sense, but this would somewhat defeat the safe operating area protection.

[Kleinstein]

The typical solution with higher voltages is to use the floating regulator. This way the regulator parts sees little of the full voltage. It is also reasonable easy to implement some kind of cascode fast tap switching to reduce the voltage at the transistors.  One could also consider higher voltage MOSFETs instead of BJTs - they seem to be a bit more tolerant in teh SOA than BJTs. Modern low voltage MOSFETs are however even more limited in the SOA. The floating regulator circuit is actually very flexible and could work with little change with a MOSFET or darlington BJT.

SOA protection can also be added the discrete way, though sometimes it can be simpler to have more power transistors and keep the protection simpler (e.g. no taking into account a variable input voltage).

For really high voltages vaccum tubes are also still an option. They are quit robust and unlike transistors they don't have the tendency to fail short.
For high power there is also the option to combine a switched mode regulator for the coarse part with linear regulator for fast response and low ripple.

[David Hess]

The typical solution with higher voltages is to use the floating regulator. This way the regulator parts sees little of the full voltage.

A floating regulator is not a solution when the output can operate at low voltage or be shorted, because the floating regulator then must sustain the entire input-to-output voltage difference, in which case a floating regulator is not required.  This is the same problem with a 317/337 driver stage and it plagues floating regulators when short circuit protection is required.

It is also reasonable easy to implement some kind of cascode fast tap switching to reduce the voltage at the transistors.

I always worry that the input decoupling capacitance located between the cascode and pass element will discharge enough energy through the pass element to damage or destroy it.  Some designs are not stable without it.

One could also consider higher voltage MOSFETs instead of BJTs - they seem to be a bit more tolerant in teh SOA than BJTs. Modern low voltage MOSFETs are however even more limited in the SOA.

I looked into this a couple years ago and except for linear rated MOSFETs, power bipolar transistors had about the same capability at high collector/drain voltages.  The big advantage that power MOSFETs seemed to have is that they are available with much larger die sizes and current ratings in larger packages.  Bipolar transistors seem to have never made the jump to newer packages which are larger.

[Kleinstein]

With the floating regulator only the power transistors have to withstand the full voltage. The control circuitry just gets it's own fixed power and controls the power transistor in collector circuit. So the control part can be low voltage (like 5-12 V), even if the output voltage goes high, like 500 V.

Some cascaded version may want extra capacitance between the stages, but not all. For a short time the power transistors can usually survive higher power in the SOA curve. That should be enough for a little capacitance of some 10s or 100s of nF if really needed.

I see more of an issue with relay tap switching and frequent switching that may cause sticking contacts from the high peak current to charge the rather large main filter capacitor to a high voltage. Switching DC with a relay is also not that great, as the sparc may stay active quite some time.

[Jwillis]

Theirs no issue tap switching for relays as long as it's done on the AC side before the rectifier. You want a relay that handles the peak current of course. But keep in mind that the rating of the relay will be the AC rating and not the DC rating. So in a case of say 50VAC at at 20A is well below the maximum rating of a 250VAC 30A relay. Also AC is far less destructive to relay contacts than DC. This is why there are 2 different ratings for a relay, AC and DC. The other thing is the bulk capacitors are not switched to the full high voltage instantly but instead by increments of voltage depending on the number of taps.
So if your first tap is relaxed at 0-10V at startup the capacitors only charge to 10V. Switch up to the next tap will charge to the next voltage like 20V and so on. So each time a tap is activated the charge will only be 10V per tap. Same thing happens when you reduce the voltage tapping down. So the energy at 10V 20A per tap at activation on a 250VAC  30A relay is less than 1/25 the power capability of the relay. The relays will last for a long time if used before the rectifier.
I due suggest for anything above 10A that a soft start is used for that initial startup.

But in the case of the OP the requirements are only 50V at 3A .